Non-functional regression : A new challenge for neural networks
- Vamplew, Peter, Dazeley, Richard, Foale, Cameron, Choudhury, Tanveer
- Authors: Vamplew, Peter , Dazeley, Richard , Foale, Cameron , Choudhury, Tanveer
- Date: 2018
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 314, no. (2018), p. 326-335
- Full Text:
- Reviewed:
- Description: This work identifies an important, previously unaddressed issue for regression based on neural networks – learning to accurately approximate problems where the output is not a function of the input (i.e. where the number of outputs required varies across input space). Such non-functional regression problems arise in a number of applications, and can not be adequately handled by existing neural network algorithms. To demonstrate the benefits possible from directly addressing non-functional regression, this paper proposes the first neural algorithm to do so – an extension of the Resource Allocating Network (RAN) which adds additional output neurons to the network structure during training. This new algorithm, called the Resource Allocating Network with Varying Output Cardinality (RANVOC), is demonstrated to be capable of learning to perform non-functional regression, on both artificially constructed data and also on the real-world task of specifying parameter settings for a plasma-spray process. Importantly RANVOC is shown to outperform not just the original RAN algorithm, but also the best possible error rates achievable by any functional form of regression.
- Authors: Vamplew, Peter , Dazeley, Richard , Foale, Cameron , Choudhury, Tanveer
- Date: 2018
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 314, no. (2018), p. 326-335
- Full Text:
- Reviewed:
- Description: This work identifies an important, previously unaddressed issue for regression based on neural networks – learning to accurately approximate problems where the output is not a function of the input (i.e. where the number of outputs required varies across input space). Such non-functional regression problems arise in a number of applications, and can not be adequately handled by existing neural network algorithms. To demonstrate the benefits possible from directly addressing non-functional regression, this paper proposes the first neural algorithm to do so – an extension of the Resource Allocating Network (RAN) which adds additional output neurons to the network structure during training. This new algorithm, called the Resource Allocating Network with Varying Output Cardinality (RANVOC), is demonstrated to be capable of learning to perform non-functional regression, on both artificially constructed data and also on the real-world task of specifying parameter settings for a plasma-spray process. Importantly RANVOC is shown to outperform not just the original RAN algorithm, but also the best possible error rates achievable by any functional form of regression.
Adaptive weighted non-parametric background model for efficient video coding
- Chakraborty, Subrata, Paul, Manoranjan, Murshed, Manzur, Ali, Mortuza
- Authors: Chakraborty, Subrata , Paul, Manoranjan , Murshed, Manzur , Ali, Mortuza
- Date: 2017
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 226, no. (2017), p. 35-45
- Full Text:
- Reviewed:
- Description: Dynamic background frame based video coding using mixture of Gaussian (MoG) based background modelling has achieved better rate distortion performance compared to the H.264 standard. However, they suffer from high computation time, low coding efficiency for dynamic videos, and prior knowledge requirement of video content. In this paper, we introduce the application of the non-parametric (NP) background modelling approach for video coding domain. We present a novel background modelling technique, called weighted non-parametric (WNP) which balances the historical trend and the recent value of the pixel intensities adaptively based on the content and characteristics of any particular video. WNP is successfully embedded into the latest HEVC video coding standard for better rate-distortion performance. Moreover, a novel scene adaptive non-parametric (SANP) technique is also developed to handle video sequences with high dynamic background. Being non-parametric, the proposed techniques naturally exhibit superior performance in dynamic background modelling without a priori knowledge of video data distribution.
- Authors: Chakraborty, Subrata , Paul, Manoranjan , Murshed, Manzur , Ali, Mortuza
- Date: 2017
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 226, no. (2017), p. 35-45
- Full Text:
- Reviewed:
- Description: Dynamic background frame based video coding using mixture of Gaussian (MoG) based background modelling has achieved better rate distortion performance compared to the H.264 standard. However, they suffer from high computation time, low coding efficiency for dynamic videos, and prior knowledge requirement of video content. In this paper, we introduce the application of the non-parametric (NP) background modelling approach for video coding domain. We present a novel background modelling technique, called weighted non-parametric (WNP) which balances the historical trend and the recent value of the pixel intensities adaptively based on the content and characteristics of any particular video. WNP is successfully embedded into the latest HEVC video coding standard for better rate-distortion performance. Moreover, a novel scene adaptive non-parametric (SANP) technique is also developed to handle video sequences with high dynamic background. Being non-parametric, the proposed techniques naturally exhibit superior performance in dynamic background modelling without a priori knowledge of video data distribution.
Softmax exploration strategies for multiobjective reinforcement learning
- Vamplew, Peter, Dazeley, Richard, Foale, Cameron
- Authors: Vamplew, Peter , Dazeley, Richard , Foale, Cameron
- Date: 2017
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 263, no. (2017), p. 74-86
- Full Text:
- Reviewed:
- Description: Despite growing interest over recent years in applying reinforcement learning to multiobjective problems, there has been little research into the applicability and effectiveness of exploration strategies within the multiobjective context. This work considers several widely-used approaches to exploration from the single-objective reinforcement learning literature, and examines their incorporation into multiobjective Q-learning. In particular this paper proposes two novel approaches which extend the softmax operator to work with vector-valued rewards. The performance of these exploration strategies is evaluated across a set of benchmark environments. Issues arising from the multiobjective formulation of these benchmarks which impact on the performance of the exploration strategies are identified. It is shown that of the techniques considered, the combination of the novel softmax–epsilon exploration with optimistic initialisation provides the most effective trade-off between exploration and exploitation.
- Authors: Vamplew, Peter , Dazeley, Richard , Foale, Cameron
- Date: 2017
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 263, no. (2017), p. 74-86
- Full Text:
- Reviewed:
- Description: Despite growing interest over recent years in applying reinforcement learning to multiobjective problems, there has been little research into the applicability and effectiveness of exploration strategies within the multiobjective context. This work considers several widely-used approaches to exploration from the single-objective reinforcement learning literature, and examines their incorporation into multiobjective Q-learning. In particular this paper proposes two novel approaches which extend the softmax operator to work with vector-valued rewards. The performance of these exploration strategies is evaluated across a set of benchmark environments. Issues arising from the multiobjective formulation of these benchmarks which impact on the performance of the exploration strategies are identified. It is shown that of the techniques considered, the combination of the novel softmax–epsilon exploration with optimistic initialisation provides the most effective trade-off between exploration and exploitation.
Steering approaches to Pareto-optimal multiobjective reinforcement learning
- Vamplew, Peter, Issabekov, Rustam, Dazeley, Richard, Foale, Cameron, Berry, Adam, Moore, Tim, Creighton, Douglas
- Authors: Vamplew, Peter , Issabekov, Rustam , Dazeley, Richard , Foale, Cameron , Berry, Adam , Moore, Tim , Creighton, Douglas
- Date: 2017
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 263, no. (2017), p. 26-38
- Full Text:
- Reviewed:
- Description: For reinforcement learning tasks with multiple objectives, it may be advantageous to learn stochastic or non-stationary policies. This paper investigates two novel algorithms for learning non-stationary policies which produce Pareto-optimal behaviour (w-steering and Q-steering), by extending prior work based on the concept of geometric steering. Empirical results demonstrate that both new algorithms offer substantial performance improvements over stationary deterministic policies, while Q-steering significantly outperforms w-steering when the agent has no information about recurrent states within the environment. It is further demonstrated that Q-steering can be used interactively by providing a human decision-maker with a visualisation of the Pareto front and allowing them to adjust the agent’s target point during learning. To demonstrate broader applicability, the use of Q-steering in combination with function approximation is also illustrated on a task involving control of local battery storage for a residential solar power system.
- Authors: Vamplew, Peter , Issabekov, Rustam , Dazeley, Richard , Foale, Cameron , Berry, Adam , Moore, Tim , Creighton, Douglas
- Date: 2017
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 263, no. (2017), p. 26-38
- Full Text:
- Reviewed:
- Description: For reinforcement learning tasks with multiple objectives, it may be advantageous to learn stochastic or non-stationary policies. This paper investigates two novel algorithms for learning non-stationary policies which produce Pareto-optimal behaviour (w-steering and Q-steering), by extending prior work based on the concept of geometric steering. Empirical results demonstrate that both new algorithms offer substantial performance improvements over stationary deterministic policies, while Q-steering significantly outperforms w-steering when the agent has no information about recurrent states within the environment. It is further demonstrated that Q-steering can be used interactively by providing a human decision-maker with a visualisation of the Pareto front and allowing them to adjust the agent’s target point during learning. To demonstrate broader applicability, the use of Q-steering in combination with function approximation is also illustrated on a task involving control of local battery storage for a residential solar power system.
Discrete state transition algorithm for unconstrained integer optimization problems
- Zhou, Xiaojun, Gao, David, Yang, Chunhua, Gui, Weihua
- Authors: Zhou, Xiaojun , Gao, David , Yang, Chunhua , Gui, Weihua
- Date: 2016
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 173, no. (2016), p. 864-874
- Full Text:
- Reviewed:
- Description: A recently new intelligent optimization algorithm called discrete state transition algorithm is considered in this study, for solving unconstrained integer optimization problems. Firstly, some key elements for discrete state transition algorithm are summarized to guide its well development. Several intelligent operators are designed for local exploitation and global exploration. Then, a dynamic adjustment strategy "risk and restoration in probability" is proposed to capture global solutions with high probability. Finally, numerical experiments are carried out to test the performance of the proposed algorithm compared with other heuristics, and they show that the similar intelligent operators can be applied to ranging from traveling salesman problem, boolean integer programming, to discrete value selection problem, which indicates the adaptability and flexibility of the proposed intelligent elements. (C) 2015 Elsevier B.V. All rights reserved.
- Authors: Zhou, Xiaojun , Gao, David , Yang, Chunhua , Gui, Weihua
- Date: 2016
- Type: Text , Journal article
- Relation: Neurocomputing Vol. 173, no. (2016), p. 864-874
- Full Text:
- Reviewed:
- Description: A recently new intelligent optimization algorithm called discrete state transition algorithm is considered in this study, for solving unconstrained integer optimization problems. Firstly, some key elements for discrete state transition algorithm are summarized to guide its well development. Several intelligent operators are designed for local exploitation and global exploration. Then, a dynamic adjustment strategy "risk and restoration in probability" is proposed to capture global solutions with high probability. Finally, numerical experiments are carried out to test the performance of the proposed algorithm compared with other heuristics, and they show that the similar intelligent operators can be applied to ranging from traveling salesman problem, boolean integer programming, to discrete value selection problem, which indicates the adaptability and flexibility of the proposed intelligent elements. (C) 2015 Elsevier B.V. All rights reserved.
Visual perceptual and handwriting skills in children with developmental coordination disorder
- Prunty, Mellissa, Barnett, Anna, Wilmut, Kate, Plumb, Mandy
- Authors: Prunty, Mellissa , Barnett, Anna , Wilmut, Kate , Plumb, Mandy
- Date: 2016
- Type: Text , Journal article
- Relation: Human Movement Science Vol. 49, no. (2016), p. 54-65
- Full Text:
- Reviewed:
- Description: Objective: Children with Developmental Coordination Disorder demonstrate a lack of automaticity in handwriting as measured by pauses during writing. Deficits in visual perception have been proposed in the literature as underlying mechanisms of handwriting difficulties in children with DCD. The aim of this study was to examine whether correlations exist between measures of visual perception and visual motor integration with measures of the handwriting product and process in children with DCD. Method: The performance of twenty-eight 8-14 year-old children who met the DSM-5 criteria for DCD was compared with 28 typically developing (TD) age and gender-matched controls. The children completed the Developmental Test of Visual Motor Integration (VMI) and the Test of Visual Perceptual Skills (TVPS). Group comparisons were made, correlations were conducted between the visual perceptual measures and handwriting measures and the sensitivity and specificity examined. Results: The DCD group performed below the TD group on the VMI and TVPS. There were no significant correlations between the VMI or TVPS and any of the handwriting measures in the DCD group. In addition, both tests demonstrated low sensitivity. Conclusion: Clinicians should execute caution in using visual perceptual measures to inform them about handwriting skill in children with DCD. © 2016 The Authors.
- Authors: Prunty, Mellissa , Barnett, Anna , Wilmut, Kate , Plumb, Mandy
- Date: 2016
- Type: Text , Journal article
- Relation: Human Movement Science Vol. 49, no. (2016), p. 54-65
- Full Text:
- Reviewed:
- Description: Objective: Children with Developmental Coordination Disorder demonstrate a lack of automaticity in handwriting as measured by pauses during writing. Deficits in visual perception have been proposed in the literature as underlying mechanisms of handwriting difficulties in children with DCD. The aim of this study was to examine whether correlations exist between measures of visual perception and visual motor integration with measures of the handwriting product and process in children with DCD. Method: The performance of twenty-eight 8-14 year-old children who met the DSM-5 criteria for DCD was compared with 28 typically developing (TD) age and gender-matched controls. The children completed the Developmental Test of Visual Motor Integration (VMI) and the Test of Visual Perceptual Skills (TVPS). Group comparisons were made, correlations were conducted between the visual perceptual measures and handwriting measures and the sensitivity and specificity examined. Results: The DCD group performed below the TD group on the VMI and TVPS. There were no significant correlations between the VMI or TVPS and any of the handwriting measures in the DCD group. In addition, both tests demonstrated low sensitivity. Conclusion: Clinicians should execute caution in using visual perceptual measures to inform them about handwriting skill in children with DCD. © 2016 The Authors.
Understanding safety management system applicability in community sport
- Donaldson, Alex, Borys, David, Finch, Caroline
- Authors: Donaldson, Alex , Borys, David , Finch, Caroline
- Date: 2013
- Type: Text , Journal article
- Relation: Safety Science Vol. 60, no. (2013), p. 95-104
- Relation: http://purl.org/au-research/grants/nhmrc/565907
- Full Text:
- Reviewed:
- Description: Despite recent interest in understanding the implementation context for sports injury prevention interventions, little research attention has been paid to the management structures and processes of community sporting organisations. This study developed expert consensus about the importance of Occupational Health and Safety (OHS) setting-related safety management system (SMS) principles and performance indicators in the context of Australian community sporting organizations, and the feasibility of these organisations meeting the requirements for the SMS performance indicators. Twenty-nine sports injury prevention, community sports administration and OHS SMS experts participated in a three-round online Delphi study by rating the importance of 64 SMS performance indicators categorised under the five principles of Commitment and Policy; Planning; Implementation; Measurement and Evaluation; and Review and Improvement. Overall, consensus agreement - define as rated 'essential' or 'very important' on a five-point scale by ≥75% of the participants in Round 3 - was reached for 57 performance indicators. Ten (15%) performance indicators were rated as 'very difficult' or 'relatively difficult', and six (9%) were rated as 'very easy' or 'relatively easy' on a four-point scale, by ≥75% of participants. This research suggests that the guiding principles and associated performance indicators that underpin OHS safety management systems in the workplace are very relevant and applicable to community sporting organisations in Australia. However, considerable work is required to build organisational capacity to be able to develop and implement meaningfully and useful SMSs to prevent sports injuries in the most common setting in which they occur. © 2013 Elsevier Ltd. Funded by NHMRC.
- Description: 2003011206
- Authors: Donaldson, Alex , Borys, David , Finch, Caroline
- Date: 2013
- Type: Text , Journal article
- Relation: Safety Science Vol. 60, no. (2013), p. 95-104
- Relation: http://purl.org/au-research/grants/nhmrc/565907
- Full Text:
- Reviewed:
- Description: Despite recent interest in understanding the implementation context for sports injury prevention interventions, little research attention has been paid to the management structures and processes of community sporting organisations. This study developed expert consensus about the importance of Occupational Health and Safety (OHS) setting-related safety management system (SMS) principles and performance indicators in the context of Australian community sporting organizations, and the feasibility of these organisations meeting the requirements for the SMS performance indicators. Twenty-nine sports injury prevention, community sports administration and OHS SMS experts participated in a three-round online Delphi study by rating the importance of 64 SMS performance indicators categorised under the five principles of Commitment and Policy; Planning; Implementation; Measurement and Evaluation; and Review and Improvement. Overall, consensus agreement - define as rated 'essential' or 'very important' on a five-point scale by ≥75% of the participants in Round 3 - was reached for 57 performance indicators. Ten (15%) performance indicators were rated as 'very difficult' or 'relatively difficult', and six (9%) were rated as 'very easy' or 'relatively easy' on a four-point scale, by ≥75% of participants. This research suggests that the guiding principles and associated performance indicators that underpin OHS safety management systems in the workplace are very relevant and applicable to community sporting organisations in Australia. However, considerable work is required to build organisational capacity to be able to develop and implement meaningfully and useful SMSs to prevent sports injuries in the most common setting in which they occur. © 2013 Elsevier Ltd. Funded by NHMRC.
- Description: 2003011206
Working to rule or working safely? Part 2 : The management of safety rules and procedures
- Authors: Hale, Andrew , Borys, David
- Date: 2012
- Type: Text , Journal article
- Relation: Safety Science Vol.55, no. (2012), p.54-59
- Full Text:
- Reviewed:
- Description: Part 1, the companion paper to this paper () reviews the literature from 1986 on the management of those safety rules and procedures which relate to the workplace level in organisations. It contrasts two different paradigms of how work rules and their development and use are perceived and managed. The first is a top-down classical, rational approach in which rules are seen as static, comprehensive limits of freedom of choice, imposed on operators at the sharp end and violations are seen as negative behaviour to be suppressed. The second is a bottom-up constructivist view of rules as dynamic, local, situated constructions of operators as experts, where competence is seen to a great extent as the ability to adapt rules to the diversity of reality. That paper explores the research underlying and illustrating these two paradigms. In this second paper we draw on that literature study to propose a framework of rule management which attempts to draw the lessons from both paradigms. It places the monitoring and adaptation of rules central to its management process and emphasises the need for participation of the intended rule followers in the processes of rule-making, but more importantly in keeping those rules alive and up to date in a process of regular and explicit dialogue with first-line supervision, and through them with the technical, safety and legal experts on the system functioning. The framework is proposed for testing in the field as a benchmark for good practice. © 2012 Elsevier Ltd. All rights reserved.
- Authors: Hale, Andrew , Borys, David
- Date: 2012
- Type: Text , Journal article
- Relation: Safety Science Vol.55, no. (2012), p.54-59
- Full Text:
- Reviewed:
- Description: Part 1, the companion paper to this paper () reviews the literature from 1986 on the management of those safety rules and procedures which relate to the workplace level in organisations. It contrasts two different paradigms of how work rules and their development and use are perceived and managed. The first is a top-down classical, rational approach in which rules are seen as static, comprehensive limits of freedom of choice, imposed on operators at the sharp end and violations are seen as negative behaviour to be suppressed. The second is a bottom-up constructivist view of rules as dynamic, local, situated constructions of operators as experts, where competence is seen to a great extent as the ability to adapt rules to the diversity of reality. That paper explores the research underlying and illustrating these two paradigms. In this second paper we draw on that literature study to propose a framework of rule management which attempts to draw the lessons from both paradigms. It places the monitoring and adaptation of rules central to its management process and emphasises the need for participation of the intended rule followers in the processes of rule-making, but more importantly in keeping those rules alive and up to date in a process of regular and explicit dialogue with first-line supervision, and through them with the technical, safety and legal experts on the system functioning. The framework is proposed for testing in the field as a benchmark for good practice. © 2012 Elsevier Ltd. All rights reserved.
Working to rule, or working safely? Part 1 : A state of the art review
- Authors: Hale, Andrew , Borys, David
- Date: 2012
- Type: Text , Journal article
- Relation: Safety Science Vol.55, no. June (2013), p.207-221
- Full Text:
- Reviewed:
- Description: The paper reviews the literature from 1986 on the management of those safety rules and procedures which relate to the workplace level in organisations. It contrasts two different paradigms of how rules and their development and use are perceived and managed. The first is a top-down classical, rational approach in which rules are seen as static, comprehensive limits of freedom of choice, imposed on operators at the sharp end and violations are seen as negative behaviour to be suppressed. The second is a bottom-up constructivist view of rules as dynamic, local, situated constructions of operators as experts, where competence is seen to a great extent as the ability to adapt rules to the diversity of reality. The paper explores the research underlying and illustrating these two paradigms, drawn from psychology, sociology and ethnography, organisational studies and behavioural economics. In a separate paper following on from this review (Hale and Borys, this issue http://www.sciencedirect.com/science/article/pii/S0925753512001312#b0285) the authors propose a framework of rule management which attempts to draw the lessons from both paradigms. It places the monitoring and adaptation of rules central to its management process. © 2012 Elsevier Ltd. All rights reserved.
- Authors: Hale, Andrew , Borys, David
- Date: 2012
- Type: Text , Journal article
- Relation: Safety Science Vol.55, no. June (2013), p.207-221
- Full Text:
- Reviewed:
- Description: The paper reviews the literature from 1986 on the management of those safety rules and procedures which relate to the workplace level in organisations. It contrasts two different paradigms of how rules and their development and use are perceived and managed. The first is a top-down classical, rational approach in which rules are seen as static, comprehensive limits of freedom of choice, imposed on operators at the sharp end and violations are seen as negative behaviour to be suppressed. The second is a bottom-up constructivist view of rules as dynamic, local, situated constructions of operators as experts, where competence is seen to a great extent as the ability to adapt rules to the diversity of reality. The paper explores the research underlying and illustrating these two paradigms, drawn from psychology, sociology and ethnography, organisational studies and behavioural economics. In a separate paper following on from this review (Hale and Borys, this issue http://www.sciencedirect.com/science/article/pii/S0925753512001312#b0285) the authors propose a framework of rule management which attempts to draw the lessons from both paradigms. It places the monitoring and adaptation of rules central to its management process. © 2012 Elsevier Ltd. All rights reserved.
- «
- ‹
- 1
- ›
- »